Can anyone recommend papers that model how long temporary price impact last when you buy / sell a trade? This would fall under the TCA realm (Trade Cost Analysis).
The reference papers for market impact decay are:
The first one uses a database of "cash trades" (ie trade with no specific decision investment), the second one uses a database of the decisions of an hedge fund, and the third one uses a database of institutional decisions.
They have the same conclusions:
permanent impact is linked to the investment decision: if you decided to buy to invest and you are a good portfolio manager, the price will ultimately go up. It is not really linked to your impact (even if you would not have traded, the price will have gone up).
temporary impact decay has two scales: a fast intraday one, and a slow (12 to 20 days) one. This impact reverts to zero.
Point (1) does not mean permanent impact is fully independent from buying / selling pressure. It can come from the cumulative pressure of all participants, it is a mean field effect: these three papers just say an isolated investor / trader cannot move permanently the price.
If you are not convinced by the upper three papers, you can have a look at this one:
It identifies market impact and decay on trades from the XVIIIth century on Amsterdam stock exchange!
You might want to measure the resiliency of a limit order book.